A cross-sectional survey in urban Delhi found a strong positive association between coffee consumption and myocardial infarction. However, when the analysis was adjusted for cigarette smoking, the association became weak and non-significant. Which of the following best explains this finding?

Explanation

## The Scenario Observations: - Raw association: Coffee ↔ MI (strong, positive) - After adjustment for smoking: Coffee ↔ MI (weak, non-significant) - Smoking is the variable that, when controlled, eliminates the association ## Criteria for Confounding **Key Point:** A variable is a confounder if and only if: 1. It is associated with the exposure (coffee consumption) 2. It is associated with the outcome (MI) 3. It is NOT on the causal pathway 4. Adjustment for it changes the estimate of association ## Why Smoking Is the Confounder **High-Yield:** Smokers are more likely to drink coffee AND more likely to have MI. When we adjust for smoking, we remove this spurious association. ``` Smoking ↙ ↘ Coffee MI ``` The apparent coffee-MI link is explained entirely by smoking, not by coffee itself. ## Adjustment and Stratification **Clinical Pearl:** When adjustment for a variable eliminates or substantially reduces an association, that variable is almost certainly a confounder. This is how confounders are identified and controlled in observational studies. ## Comparison with Other Explanations | Explanation | Why Not Correct | |-------------|------------------| | Coffee causes MI (option 0) | If true, adjustment would NOT eliminate the association; a causal pathway would persist | | Information bias (option 2) | Affects measurement accuracy, not the relationship between variables; adjustment would not resolve it | | Selection bias (option 3) | Affects who enters the study; adjustment for smoking does not address recruitment issues | **Mnemonic:** **CHANGE** — Confounding is Halted After Neutral Grouping/stratification by the Explanatory variable (the confounder).